Capacitor Discharge
Welder
Part 2
By Phil Prosser
This Capacitor Discharge Welder
has been carefully designed to
deliver just the right amount of weld
energy each time. When completed,
it makes a neat package that’s easy
to build and safe to use, so long
as you follow our advice. Having
described how it works last month,
let’s get into making it.
safe and low voltage
T
he Capacitor Discharge
Welder comprises three main
electronic modules: the Power
Supply, which is responsible for
charging the capacitors; the Controller
Module, which determines when voltage is applied across the welding tips;
and an Energy Storage Module bank,
typically made from around 10 modules joined to a common pair of bus
bars, that hold the storage capacitors
and MOSFETs.
Because of this modular approach,
not only can you scale the system to
meet your needs, but the PCB cost is
kept down, and assembly is relatively
straightforward. You build and test the
modules, assemble them into the case,
make the welding tips and cables, and
finally wire it all up.
Construction
The first step in building the CD
Welder is to assemble one Power
Supply Module, one Controller Module and several Energy Storage Modules (ESMs). Each is built on a different PCB, but all the PCBs are the same
size at 150 x 42.5mm.
We’ll start with the Power Supply
Module. Its PCB is coded 29103221 and
Fig.6 is its overlay diagram, which shows
the parts’ location mount on the board.
Start by soldering the sole SMD
ceramic capacitor (100nF) near the
MC34167 regulator IC. Next, mount the
Safety warning
Capacitor Discharge Welding works by generating extremely high current pulses, and
consequently, strong magnetic fields. Do not build or use this project if you have a
pacemaker or similar sensitive device.
This device can generate sparks and heat. Users must wear appropriate personal
protective equipment such as AS/NZS 1337.1, DIN 169 Shade 3 welding glasses.
These provide mechanical and IR/UV protection.
26
Practical Electronics | April | 2023
INA282 current sense amplifier, which
comes in an SMD package (SOIC).
Watch its orientation; make sure its
pin 1 is facing as shown before soldering its pins and then check for bridges.
Follow with all the resistors and
diodes (except for diode D1) with the
diode cathode stripes facing as shown,
leaving the taller shunt resistor until
last. There are three different diode
types used: 1N4148, 1N4004 and zener,
so don’t get them mixed up.
Pay attention to the two different
resistor value options shown in Fig.6.
If you are using a DC power supply that
can deliver at least 5A, you can use the
values shown for 5A charging. Otherwise, stick with 2A charging.
Now install the sole transistor, facing
as shown, then all the capacitors. Many
of the latter are not polarised, but for
those which are polarised (the electrolytics), these all have the longer positive leads going to pads on the righthand side. Note that while you could
use 100nF MKT capacitors, multi-layer
ceramics will also work.
Next come the connectors. There
are two screw terminals, a polarised
header for the Charge LED and a 2x5
pin header to connect to the other
modules. Make sure the screw terminal wire entries face the outside of the
board as shown.
Mount the 6TQ045-M3 diode (D1)
close to the board by pushing it down
fully before soldering and trimming its
leads. Also install the fuse clips (with
the tabs towards the outside) and fuse,
the LM358 op amp and 10kW linear
voltage control potentiometer.
Now fit the LM7815 regulator and
attach a small flag heatsink using a
machine screw, shakeproof washer and
nut as it gets warm during operation.
Mount the 220µH toroidal inductor on the board, then finally the
MC34167 switch-mode regulator.
This also requires a small heatsink
such as Altronics H0625 with an
insulating bush and silicone pad.
Hold this all together using an M3
machine screw, star washer and nut
in the usual manner.
The CD Welder fully assembled
and ready to be used in anger (or
calmly, it’s up to you).
Control board
The Controller PCB is coded 29103222
– refer to Fig.7.
Start by installing all the resistors
and diodes, checking that the diodes
are the right way around, then follow
with the four NE555 timer chips, with
their pin 1 notches/dots to the left.
Next, fit all the ceramic MKT and
electrolytic capacitors. Note the two
different types of 1µF capacitor as well
as different types of 220nF capacitors.
The electrolytics have longer leads for
their positive connections, and these
go to the side marked + on the overlay.
Now mount the small transistor, facing as shown, followed by the 100kW
linear potentiometer and the 2-way and
10-way headers.
If you want to make the controller
switchable for two pulses, make a cable
with a switch at one end and a header
plug on the other so that it can plug
into CON8. Alternatively, you could
install a jumper on CON8 and fix this
setting, as we did.
Energy Storage Modules
The ESM boards are coded 29103223,
and the components are mounted as
shown in Fig.8 and Fig.9. Presumably
by now you will have figured out how
many you need to build and obtained
the appropriate capacitors. Generally,
there are three caps per board, but
some of the recommended configurations use two. In this case, fit the two
closest to the headers.
Start by fitting the surface-mount
resistors and capacitors on the underside of the PCB. Make sure the 100nF
capacitors are mounted either side of
the MOSFET driver (IC8). Then solder
that driver IC, being careful not to short
Fig.6: the Power Supply board is built mainly using through-hole components. The only SMDs are IC2 and one 100nF
capacitor near IC1, so fit those first. Watch the orientations of IC2, IC3, the diodes, electrolytic capacitors, REG1 and the
terminal blocks.
Practical Electronics | April | 2023
27
any leads (you can clean up any bridges
using flux paste and solder wick).
Next, mount the RFN20NS flyback
diode (D9) to the PCB. It’s easier if you
spread a thin layer of flux paste on all
its pads first. You will want to get a
good lot of heat into the PCB; start by
tacking down the two anode leads, then
solder the main body of the diode. This
will not dissipate much power, but you
want a good solder joint here.
Then fit the two MOSFETs, keeping
their leads short. Their metal tabs face
away from the capacitors, and their
source and drain pins connect to copper fills. These junctions will see very
high current pulses, so be sure to get
these properly hot when soldering to
form nice-looking fillets.
Now mount the 2x5 control header,
the terminal block and finally, the
capacitors. Make sure their positive
sides go in the direction indicated, and
the negative side stripes face away from
this. (Reversed capacitors will likely
lead to an earth-shattering kaboom!)
Repeat the ESM assembly until you
have enough of these modules, and are
ready to test them and then proceed to
final assembly.
Testing
Start by testing the modules individually, beginning with the Power Supply
Module. To start with, solder the leads
of one LED to a length of light-duty
twin-lead cable (eg, two wires stripped
from ribbon cable) and solder/crimp
the other end into a pluggable header,
and connect this to CON3, the charge
LED header. Make sure the anode (longer LED lead) goes to pin 1.
Connect the Power Supply board to
a DC voltage source of at least 25V –
up to 35V is acceptable. Make sure you
have set the current limit (2A or 5A)
to match your supply. Set your DVM
to a DC volts range and put a 5W 82W
resistor across CON2, ‘Power Output’.
Apply power and check the following:
n
The output of the LM7815 is 15V
±0.25V. Its output is accessible on
pin 2 of CON4, the control header.
If not, check that it is the right way
around and there are no shorts.
n
Check pin 1 of CON2, the ‘Power Out’
connector, is between 2V and 25V.
Also check that this can be controlled
using potentiometer VR1. If this is not
working, check the following:
n
Check the INA282 (IC2) is in the right
way around.
n
Verify that the 82W test resistor is
connected correctly (eg, measure
the resistance across the terminals
of CON2).
n
Check the MC34167 is oscillating;
there will be a 72kHz signal at pin 2.
n
Check D1 is in the right way around.
n
Check that the feedback pin 1 of the
MC34167 has about 5.05V on it. If
not, verify that the LM358 op amp is
operating properly. Check the voltages at its power and ground pins
(pins 8 and 1, respectively), and verify that the voltage at input pin 5 is
an appropriate fraction of the output
voltage, and that pin 7 is an amplified
version of this. Check that diodes D4
and D5 are in the right way around.
n
Assuming that’s working, put an
ammeter on its 10A range across the
terminals of CON2 and check that the
current is close to the expected 2A
or 5A. If not, look for problems near
the INA282 (IC2).
Testing the Controller
To test the controller, ideally, you will
need an oscilloscope. Make a 10-way
IDC lead to connect the Power Supply module to the Controller module,
ensuring that pin 1 connects to pin 1.
Apply power and check the following:
n
Each NE555 chip has 15V at its pin 8.
n
The base of transistor Q1 is pulled up
to within 0.6V of the 15V rail, turning it off.
n
The TRIGGER output of IC6 (pin 3)
is close to 0V
The next part is easiest if you assemble the foot pedal trigger by extending
the existing lead with the two-metre
length of microphone cable. You can
simply snip off the screen wires as they
are not required; just use the two internal conductors, then add liberal layers
of heatshrink to protect the junction.
Now temporarily soldering a length
of light-duty twin lead to the other end
(eg, stripped from spare ribbon cable)
and solder/crimp this to a polarised
header plug which connects to CON5.
Connect your oscilloscope to the
output pins (pin 3) of IC4, IC5 and
IC7. If you only have a single-channel or two-channel oscilloscope, start
with IC4 and/or IC5 and then test the
rest later.
Press the footswitch and check that
IC4 generates a pulse of about 0.1ms
and IC5 generates a pulse of about
5ms. Then check that IC7 generates a
pulse length that is controllable using
potentiometer VR2, from about 0.2ms
to over 20ms.
Next, check that the trigger output
on pin 9 of the 2x5 header (or pin 3 of
IC7) contains one or two pulses as set
by the switch/jumper on CON8.
If there are problems, check the
power supply to the NE555 ICs; there
should be 15V between pins 8 and 1
of each chip.
Verify that the trigger input (pin 2)
is being pulled low on IC4, and that
the inputs to subsequent NE555s have
a short negative-going pulse (this is
capacitively coupled, so look closely
with the scope).
Check also that the diodes are in the
right way around, that Q2 is indeed a
PNP device and that the INHIBIT line
is not pulled low by the Power Supply.
Make sure that you are happy with
the operation of the power supply and
controller modules before assembling
the CD Welder.
Testing the ESMs
To check out each Energy Storage
Module, connect one at a time to the
Controller and Charger modules. Use
medium-duty hookup wire (0.7mm
diameter copper/21AWG) such as
Altronics Cat W2261/W2260 or Jaycar Cat WH3045/WH3046 to connect
the Power Out connector on the Power
Supply board (CON2) to the Power
In connector (CON10) on the Energy
Store Module.
You’ll also need a control ribbon
cable with three 10-way IDC line sockets to connect the Power Supply, Controller board and ESM together.
Fig.7: the Control board uses all through-hole parts and assembly is straightforward. Again, be careful to orient the
diodes, electrolytic capacitors and ICs as shown.
28
Practical Electronics | April | 2023
The bus bar
layout for 10
modules,
five on
either side
of the bus
bars. The
holes at the
end of the bus
bars are drilled
and tapped for
M4 to secure the
welding leads; all the other
holes are tapped M3. We have allowed
enough length for the bus bars to protrude through
holes in the case, as we do not want any joints in these.
Connect an 82W 5W test resistor
across the ESM output using 16mm
M3 machine screws, nuts and washers. Apply power and check that the
capacitors charge and that you can
adjust the voltage using VR1.
The ‘Output -VE’ connection (right
near the edge of the PCB) will be
pulled up to the same voltage by that
82W resistor. Use an oscilloscope to
watch the voltage on that pin and
press the trigger. There is a convenient ground on the power header; we
also added a ground via on the board
between the capacitors.
After triggering, you should be able
to see the output pulled to ground in
two pulses (with dual pulse mode
on). If this does not work, use the
scope to check for the trigger pulses
on the control cable, check the +15V
rail and check that the TC1427 is
sending pulses to the MOSFET gates.
Check all cabling and the orientation
of the components.
Now swap that 82W resistor for a
0.27W 5W resistor. Repeat the test, and
check that everything works. At 25V,
this will pass close to 100A.
You will see the Charge LED come
on, especially with long pulse lengths
and high voltages. You will also feel
the 0.27W resistor get hot after several
shots. This is normal. You may blow
this resistor, so if things look odd,
check it is still 0.27W.
Bus bars
Once you’ve tested the modules, it’s
time to put them all together.
We have laid these boards out such
that they can mount back-to-back
on two 260mm-long bus bars. Fig.11
shows where to drill holes to allow M3
screws to hold pairs of modules into
common tapped holes.
Mount the modules to the bus
bar using 6mm-long M3 panhead
machine screws and star washers.
As you assemble the modules to the
bus bars, put 10mm M3 spacers, 6mm
screws and star washers between the
holes at the far end of the PCBs from
the bus bars, securing pairs of boards
to one another, stabilising the assembly. Now tighten the screws well; these
will be carrying a lot of current.
You may find another way to lay
the modules out. While it might be
possible to run machine screws right
through holes drilled in the bus bars
with nuts on the other side, we feel
that using threaded holes into the
aluminium is important to keep the
resistance down. So we strongly
advise you to take the time to tap all
these holes (aluminium is soft, and
you can use a through-tap, so it isn’t
that hard).
Cabling
We have endeavoured to keep cabling
as simple as possible. Fig.10 shows
the complete layout. We extended
the ribbon and power cable from the
Energy Store Modules to the Charge
and Control modules to suit our
application. Try not to make these
more than a few hundred millimetres long.
Fig.12 shows the layout we came up
with to fit the modules inside the case
and how most of the wiring is routed.
Note that it is necessary to cut the
Inhibit line in the ribbon cable so that
it only connects the Controller and
Power Supply modules. This is to prevent it from acting as an antenna and
picking up pulses during welding.
Fig.8 and Fig.9:
the ESM has
parts on both
sides, although
the underside
components
are limited to a
few SMDs near
the MOSFETs;
mainly, the
driver IC and
associated
passives. Fit all
those first, then
flip the board
over and solder
the remaining
components to
the top side. Be
very careful with
the electrolytic
capacitor
and MOSFET
orientations, as
putting them in
backwards would
be disastrous.
Practical Electronics | April | 2023
29
You will need to make up a cable for
the enable switch similar to the one you
made before for the charge LED. This
will plug into CON6 at one end and go
to the terminals of a toggle switch at
the other end.
Now would also be a good time
to disconnect the twin lead from the
microphone cable in the footswitch
assembly you made earlier, and instead
solder these to the microphone plug
(footswitch end) and socket specified
in the parts list last month.
In our application, we started with
300mm lengths of twin lead and
trimmed them as required.
The power connection from the
chassis DC socket to the Power Supply
board needs to be made using 5A-rated
cable; the type of wire used earlier to
connect the Power Supply to the ESMs
should be suitable.
30
While the ribbon cable connects the
output of the Power Supply to each
ESM, it is only rated at 1A per wire.
Two wires are used for power, plus two
for ground, limiting charging over the
ribbon cable to 2A.
So if you want to charge at 5A,
the IDC headers will ‘need help’.
This is the purpose of CON10 on
each ESM. You will need to wire all
those headers back to CON2 on the
Power Supply using 5A-rated cable.
We used Altronics Cat W2109 for
this job. Don’t use thicker wire if you
can avoid it, as you need to fit two
pairs into each terminal block to daisy-chain them.
For this, we cut nine 60mm lengths
plus one long length, stripped and
tinned these together and used a bit
of heatshrink to make it look tidy.
This is a little fiddly, but it is the best
approach we could come up with that
was not big or too expensive. By paralleling the ribbon cable, this heavy-duty
wire will take the majority of current
during charging.
Make sure you connect each terminal with the same polarity; otherwise,
it will short out the Power Supply!
To make the ribbon cable that connects all the modules, assuming you
have 10 ESMs, you need 12 10-way
IDC line sockets and about 610mm of
10-way ribbon cable, depending on
your layout. Fit the IDC connectors as
shown in Fig.13.
We crimped the IDC connectors
using a vice, although specific tools
are also available to do this. If using
a vice, add timber blocks or sheets on
either side of the connectors to avoid
marring them and make it less likely
to break them when squeezed.
Practical Electronics | April | 2023
As mentioned earlier, we recommend cutting the inhibit line (wire 7)
between the Power Supply Module and
the Energy Store modules. Simply slit
the ribbon cable on either side of wire
7 over a 10mm section and snip a 5mm
section from it using side cutters. This
reduces the chance of EMI pick-up.
Cables
The footswitch is our solution to keeping your hands free to weld, but you
could place a button on one of the
leads as an alternative if you wish.
The recommended footswitch comes
with a short lead, hence our earlier
instructions to extend it with about
two metres of microphone cable. Now
that you’ve added the plug and socket,
this cable should be complete.
For the all-important welding
cables, we crimped Altronics H1757B
non-insulated eyelet lugs at the Welder
end (Jaycar PT4936 is equivalent).
We were lucky and our crimping tool
worked on these, but we know from
experience that you can also solder them
(with a powerful iron) or crimp them in
a vice. We put 10mm heatshrink over the
terminal to ensure nothing shorts to it.
We made the welding handles and
tips as shown in Fig.14. These comprise a 100mm length of 10mm square
aluminium bar with a 4mm hole drilled
in the end to accept the welding cable.
Two additional M4 threaded holes
allow 6mm-long M4 screws to fix the
welding cable.
After making them, we applied
13mm heatshrink tubing over the handles to make them easier to hold and
act as strain relief for the cables.
At the welding tips, we have again
drilled 3mm holes in the end of the
handles and drilled and tapped an
M3 threaded hole to hold the tip. We
tried various copper welding tips and
feel that 3mm rod filed to a point are
pretty good.
We used small pieces of 20mm heatshrink to ensure the positive and negative welding cables remain close to one
another along the bulk of their length.
We do this to minimise the inductance
in the welding cable loop. If there is a
lot of inductance, then there will be
so much energy stored in this that the
MOSFETs have to switch, and the flyback diodes need to redirect.
Case assembly
There are many ways of packaging
this up. By avoiding mains wiring,
we don’t need to be so worried about
earthing and suchlike. We used an
Altronics H0364A case, which is just
Reproduced by arrangement with
SILICON CHIP magazine 2023.
www.siliconchip.com.au
Fig.10 (left): this shows the
required cabling for the
complete system, which is
relatively simple. You can
have more or fewer ESMs,
but six is the minimum.
All cables connect to
headers or terminal
blocks, except the optional
voltmeter we added,
which tacks onto a solder
pad that joins to the +15V
supply rail.
Fig.11 (below): to make
the bus bars, cut 10mm
square aluminium bar to
two 260mm lengths and
drill and tap M3 holes in
the locations shown. Use
kerosene or light machine
oil to lubricate the tap
and if it sticks, withdraw
it and clear out the swarf
before continuing. You
don’t want to break the
tap off in the bar.
Practical Electronics | April | 2023
31
Fig.12: this diagram shows how we mounted the modules in the recommended case and wired them up (62.5% scale).
large enough to fit all the modules.
This allows us to mount the ESM
‘bundle’ on its bus bars in the base
with the Power Supply and Controller
modules just behind the front panel,
secured to the side of the case.
The photograph of the case with
the lid off shows this arrangement
pretty clearly.
We found that the potentiometer
shafts were only just long enough
– you might find a better way of
mounting these. As our application
is stationary in the lab, we used long
tie wraps (thick cable ties) to secure
the energy store to the case and put
firm foam under the lid to hold it all
together when the lid is attached.
We folded and mounted a sheet of
Presspahn between the output bus
bars (visible in the lead photo) to
ensure that accidental shorts cannot
easily occur. Note that there is no danger here unless the ‘trigger’ footswitch
is pressed, but we do not want any
chance of accidentally firing into a
dead short. The cutting and folding
details for this are shown in Fig.15.
We cut two square holes in the front
of the case to allow the bus bars to
poke through, shown in Fig.16, along
with the other front-panel cutouts.
All controls were placed in locations
that felt convenient, and we used four
holes to fix the Presspahn sheet to the
front panel.
We found a cheap voltmeter on
eBay and decided to add this – these
are available on your favourite auction site for a few dollars if you go
looking. We will leave the selection
and integration of this to you, as there
are many choices out there, and the
wiring is pretty straightforward.
Welding!
You will need to experiment to find
the settings that work best for you. We
used flat AA and D cells to test the system out, and found that with 0.12mm
nickel strip, setting the pulse width to
maximum and voltage to about 12-14V
gave extremely solid welds.
We started with a low voltage and
increased the voltage until the welds
just stuck, which was about 8V. From
that point, we increased the voltage to
get a solid weld (in our case, at around
12V), then added a bit.
To test your welds, take pliers and try
to pull the tab off. It should be exceptionally well attached and require you
to tear the weld ‘beads’ off.
You will find the copper weld tips
wear and get dirty if you experience
arcing. Clean them up with sandpaper or a sharp knife for consistent
Fig.13: we used 610mm of ribbon
cable to connect our 12 modules as
shown here. Adjust the total length
and connector positions if you aren’t
using 10 ESMs or want to arrange
them in a different layout.
32
Practical Electronics | April | 2023
The finished Capacitor Discharge Welder, with the welding cables attached.
Fig.14: a cross-section of the welding probes we made from 10mm square aluminium bar. The welding tips are 3mm
copper rods ground to a sharp point. A close-up of one of the tips is shown adjacent to this diagram.
results. Once you have worked your
settings out, this CD Welder should
provide solid service and consistent
weld energy.
Some tips
n
We found 12-15V to be the sweet
spot for welding. While we did
install 25V capacitors, if you are
welding only light gauge battery
tabs, you will probably find that
you need to charge them no higher
than 16V. Then again, you gain a
lot of headroom for the slight cost
increase of using 25V capacitors.
n
To check the effect of weld energy,
we welded tabs to the top of a soup
can, using this as a battery surrogate. From the outside, the 15V
welds are reasonably light ‘dimples’, while with the 25V welds,
some of the tab material has clearly
Practical Electronics | April | 2023
been blown away. This was accompanied by sparks and a flash. The
photo of the inside of the can shows
that all the welds are visible, but
with significantly more damage on
the 25V welds.
n
Never short the output bus bars
directly (with a screwdriver, for
example); this will lead to dangerous arcing and quite possibly break
something expensive.
n
Always wear safety glasses.
n
Do not use welding leads with copper wider than 3.3mm in diameter
(8 Gauge) or shorter than 1m, as this
forms part of the design.
Fig.15: cut, drill and fold the
Presspahn as shown here to
make the bus bar insulator.
This ensures that the Welder
cannot be accidentally fired
with a short circuit across
the bus bars.
Holes A are 3mm in
diameter. All dimensions
are in millimetres.
33
n
Always keep the leads parallel and
never curl them into a coil. Coiling
leads will increase inductance in the
system and give the flyback diodes a
hard time.
n
Note that some plug packs have their
negative output connected to mains
earth. Be careful of these packs as the
output leads are at your weld voltage.
To illustrate the energy involved, and
potential danger, this shows the result of
placing the probes across the tab between
two AA cells. The capacitors were
charged to 15V, about 127J of energy.
A look inside a can used for testing.
This image shows the damage caused
by excessive voltage. The higher
energy welds have made holes right
through the metal.
Finally, we have a couple of spreadsheets available for download from
the April 2023 page of the PE website.
These include many of the calculations used to verify this design – see:
https://bit.ly/pe-downloads
Fig.16: the front panel
cutting diagram for
the layout used in our
prototype. This box suits
our application in the
lab, but you might be
able to come up with a
better arrangement.
34
Practical Electronics | April | 2023
